Re: Defining "flavors" of subClassOf in OWL ?

On Thu, 2003-07-10 at 11:19, Bernard Vatant wrote:
> In Mondeca ITM internal management of classes, we have a generic
> functionality somehow equivalent to the distinction between "abstract" and
> "concrete" classes in Protégé. However, the distinction is not made by
> typing the classes themselves, but by typing the class-subClass
> relationship (aka Topic Map association). This means we have two "flavors"
> of subclassing, supporting different technical treatments.
> 
> Trying to represent this in OWL, I defined two properties like
> 
> p1 rdfs:subPropertyOf rdfs:subClassOf
> p2 rdfs:subPropertyOf rdfs:subClassOf
> 
> Now if I have three classes X, Y, Z such as (X p1 Y) and (Y p2 Z)
> 
> I expect from the semantics of subPropertyOf that an OWL validator would
> infer:
> (X subClassOf Y) (Y subClassOf Z) (X subClassOf Z)

That conclusion does follow in OWL full, since...

"T-interpretations must meet several other conditions, as detailed in
the RDF semantics."
 -- http://www.w3.org/TR/owl-semantics/rdfs.html#5.2

<x,y> is in IEXT(I(rdfs:subPropertyOf)) if and only if x and y are in IP
and IEXT(x) is a subset of IEXT(y)

-- http://www.w3.org/TR/2003/WD-rdf-mt-20030123/


I'm not sure exactly what you mean by "OWL validator".


> Meaning if an ObjectProperty q has been declared e.g. of range Z, it should
> be validated if used with values in X.

I think your intuitions are in the right direction, but again,
I'm not sure what "validated" means, exactly.


> Not quite sure about it, I've tested precisely that situation on an example
> [1], and my two favourite OWL on-line validators seem to differ on the
> validation results.
> 
> http://phoebus.cs.man.ac.uk:9999/OWL/Validator seems happy with it.

That one just tells you whether a document is in one of the dialects;
every RDF document is an OWL Full document, so you're not getting
much information there. That tool doesn't find inconsistencies,
for example.

> http://owl.bbn.com/validator/ is not, and sends back a bunch of "range
> mismatch" error.

That employs a number of heuristics to find common problems.
It seems to be confused in this case.

> Is this a bug in the validator, or is it a borderline example,

Playing with the built-in vocabulary is in some sense borderline,
though it's perfectly well specified in OWL full.

>  or is it
> that subtyping "subClassOf" is altogether invalid in OWL,

only in OWL DL (and hence in OWL Lite).

>  or what?
> 
> A bottom line question is to know if that could be considered a
> recommended/neutral/bad practice.

It seems like a pretty natural way to model your situation,
and the conclusions you mention do follow...

If you're trying to get a lot of "yes, that document is squeaky
clean" checking, you might want to stay inside OWL DL;
you'll have to find a work-around for expressing what you
want to express in that case, though.

See section 3.1.3. Design for Use
http://www.w3.org/TR/owl-guide/#DesignForUse
for a discussion of some of the trade-offs.

> Thanks for your help
> 
> Bernard Vatant
> Senior Consultant
> Knowledge Engineering
> Mondeca - www.mondeca.com
> bernard.vatant@mondeca.com
> 
> [1]
> http://www.daml.org/cgi-bin/hyperdaml?http://www.mondeca.com/owl/itmex.rdf

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Thursday, 10 July 2003 17:39:39 UTC